Lunar Mass Delivery

The capture-cost inversion

Why Mass Drivers Change Which Destination Is Cheapest

A chemical vehicle pays propellant for the entire climb. A mass driver pays electricity on the track and propellant only for capture.

A chemical rocket leaving the lunar surface for LEO pays Δv for the full trip — roughly 6.3 km/s surface to low Earth orbit, counting gravity losses. A mass driver supplies the launch impulse electrically while the payload is still on the track. Once the package leaves the Moon, the economically relevant cost is capture at the destination — not total Δv from the surface.

That distinction changes which destination is cheapest by an order of magnitude. Total surface-to-destination Δv makes the destinations look different by a factor of roughly three. Capture-only Δv makes them different by roughly 26.

The mass driver analyzed in the lunar architecture leaves the track at about 2.52 km/s. Reaching EML1 or EML2 from the lunar surface requires about 2.5 km/s. The machine as designed already reaches the cislunar gateways. Everything added to push cargo below GEO — a solar-electric capture stage, propellant, months of spiral time, and a huge in-transit population crossing operational conjunction-management shells — exists only because the cargo is being pushed past GEO.

EML1 and EML2 are the natural first export terminals of a lunar industrial economy.

Capture Δv table

Capture Cost by Destination from a Lunar Mass Driver

The electrically supplied launch impulse is already paid on the surface. The relevant cost is capture at the receiving node.

What survives the track

What a Mass Driver Can and Cannot Launch

101 g sorts the cargo manifest before the first package flies.

The reference mass driver accelerates at roughly 101 g on a 3.2 km track. Shorter concepts push higher. That acceleration defines what can ride the track.

What survives: rugged radiator panels, structural beams, shielding mass, bulk regolith, sintered bricks, aluminum and titanium stock, glass, water ice in insulated slugs, oxygen in pressure vessels. Anything dense, robust, and undifferentiated.

What does not survive: delicate spacecraft, electronics assemblies, solar arrays, optical systems, crewed vehicles. A Starship-class vehicle or a Starmind satellite cannot be fired as a kit and assembled at the receiving end.

EML2 is a receiving node, not an assembly yard. Final assembly belongs where the parts are manufactured or where crews and factories already exist — at a cislunar station, in a LEO factory, or on the lunar surface.

The aiming tolerance

Muzzle-Velocity Sensitivity: Where Error Amplifies

One meter per second of launch error shifts the perigee by 220 kilometers.

Near the 2.52 km/s design point, each meter per second of muzzle-velocity error shifts the eventual Earth-perigee by roughly 220 km. The payload leaves the Moon only modestly above lunar escape velocity, so small errors in muzzle speed are amplified roughly threefold in the excess velocity that shapes the Earth-return ellipse.

The tolerance depends on destination. Direct LEO delivery requires ±1.3 m/s — about five parts in 10,000, with the dangerous error direction capable of dropping perigee into the atmosphere. GEO tolerates ±32 m/s. EML1 and EML2, which accept manifold insertion rather than aerodynamic or propulsive capture against a hard velocity target, are the most forgiving destinations.

Tight aiming is an engineering constraint on the track. Loose aiming is a property of the destination. That is another reason the first lunar exports want to stop high.

The catch problem

Who Catches: The Narrowest Dependency

The Moon can throw the mass. The unresolved question is who catches it.

The straightforward architecture is to stop cargo at EML1 or EML2 and let a reusable electric tug carry it inward. But the tug’s noble-gas propellant has no off-Earth source. Chemical alternatives push the propellant fraction toward three-quarters of departure mass. Aerobraking at 10 km/s is unsolved at the required scale. Expendable capture stages give up the reusable tug entirely.

The current concept contains an even narrower dependency. Direct delivery to low orbital shells assumes a SpaceX AI1-class vehicle can self-capture with roughly a six-percent propellant fraction. That is a spacecraft-design choice being made by SpaceX for reasons unrelated to lunar logistics. If the design changes, the arrival architecture changes with it.

No alternative self-capture stage exists at the required scale. The mass driver can be built to demonstrated engineering tolerances. The receiver — the vehicle that matches velocity with arriving cargo and delivers it to a customer orbit — does not yet have a second source.

The GEO waypoint trap

GEO Does Not Rescue the Low-Orbit Delivery Problem

GEO is cheap as a destination. It is expensive as a waypoint.

It is cheap to capture into GEO because GEO is the destination — the relative arrival velocity is low. Using GEO as an intermediate waypoint for a 2,000 km Starmind shell costs roughly 7.5 km/s in total, versus about 3.0 km/s for direct capture into the same shell. GEO is a 2.5× penalty as a route.

The cheap capture at GEO is non-transferable. The low arrival velocity is an artifact of GEO’s altitude, not a property that can be borrowed by a lower orbit. Routing cargo through GEO to reach LEO adds Δv, adds time, adds vehicles, and adds conjunction-management burden in every shell the cargo crosses on the way down.

The canister concept

Standardized Canisters: Making the Package Useful

The container itself should become useful mass.

A mass driver does not throw raw rocks. It launches standardized canisters — and the canister itself can be designed as useful product. An aluminum sleeve becomes structural stock. A glass or ceramic shell becomes shielding. A magnetic capture feature enables eddy-current or tether-based catching without propulsive rendezvous. A radiator-panel blank can be finished and deployed at the receiving node.

The canister concept eliminates a class of waste. In a conventional launch, the payload adapter and fairing are discarded. In a mass-driver architecture, the container is the first piece of delivered product. Every gram of structure that survives the track and the catch is feedstock that did not ride a rocket from Earth.

The catcher is designed as part of the launcher: a regenerative eddy-current brake, a rotating tether, a capture mesh, or a momentum-exchange station at EML1. Misses must go to a surveyed disposal zone — no persistent untracked debris.

SpaceX and the mass-driver revival

SpaceX Project TERAFAB and the Mass-Driver Roadmap

The concept is from 1976. The money is from 2026.

Gerard K. O’Neill and Henry Kolm built three mass-driver prototypes at Princeton and MIT between 1976 and the early 1980s, achieving accelerations up to 1,800 g. At that acceleration, only 160 meters of track would be needed to reach lunar escape velocity. The physics was never in question. The missing element was a vehicle that could deliver the construction equipment to the lunar surface at relevant scale.

In February 2026, Elon Musk publicly described a lunar electromagnetic launcher during an xAI all-hands meeting, and SpaceX formally placed mass drivers in its lunar roadmap as part of Project TERAFAB — a plan to relocate petawatt-scale AI compute infrastructure to the lunar surface. General Atomics filed a proposal with the Air Force Office of Scientific Research outlining how to adapt the U.S. Navy’s Electromagnetic Aircraft Launch System (EMALS) for lunar application.

Starship HLS can deliver over 100 metric tons to the lunar surface per mission. That is the construction equipment, power systems, and raw track needed to build a mass driver. The enabling logistics vehicle now exists on paper; the mass driver remains a projected capability, not a funded program with a timeline.

China’s approach

China’s Lunar Mass-Driver and ISRU Studies

China views mass drivers as central to lunar industrialization.

Chinese researchers have proposed magnetic launchers on the lunar surface to send payloads to lunar orbit or Earth-return trajectories, claiming the system could operate at roughly 10% of the cost of conventional rockets while supporting frequent automated launches. The concept fits within the broader ILRS architecture, where Chang’e-8 (targeting 2028) will demonstrate in-situ resource utilization technologies that would supply feedstock to a mass driver.

China’s sequencing is surface-first: survey the resources (Chang’e-7, ~2026), demonstrate extraction (Chang’e-8, ~2028), build the basic station (by 2035), then scale the export infrastructure. A mass driver appears in the expanded ILRS Phase 2, not in the initial build.

The ground-truth measurement

The Single Most Important Near-Term Measurement

Orbital data says the water is there. Nobody has measured how much per kilogram of regolith, in what form, or at what depth.

A mass driver can throw regolith, sintered structural elements, metals, and glass without depending on lunar water. But the full cislunar industrial loop — propellant from electrolysis, oxygen for life support, hydrogen for chemical engines, water for shielding — depends on polar ice existing in mineable quantities.

Orbital and impactor measurements from LCROSS, the LRO Diviner radiometer and neutron spectrometer, Chandrayaan-1’s Moon Mineralogy Mapper, and Korea’s Danuri ShadowCam all strongly suggest water ice is present in permanently shadowed craters at the south pole — in some form, at some concentration, in some distribution.

What is missing: direct, in-situ measurement of how much water there is per kilogram of regolith, what mineral or physical form it takes, how it is distributed laterally and with depth, and how it behaves under the heat of an extraction reactor.

Chang’e-7 (~2026) will survey the Shackleton crater region with ground-penetrating radar, a rover-mounted volatile detector, and a hopping probe that can enter permanently shadowed craters. Blue Origin’s MK1 cargo lander, targeting early 2026, will land near Shackleton to validate surface systems before carrying NASA payloads in follow-on missions. The ground-truth measurement that either confirms or falsifies the lunar-water supply chain is the single most consequential near-term data point for everything on this page.